[GPT-5.6 Sol Launches Ultrafast Mode, Speed Increased by 14x] Deep Tide TechFlow reports that on August 14, OpenAI previewed the Ultrafast mode of GPT-5.6 Sol, with output speeds reaching up to 750 tokens per second, a 14x improvement compared to previous versions. The acceleration is supported by Cerebras. Earlier this year, the two parties signed a multi-billion-dollar cooperation agreement. Initially, this service will only be available to select clients via the OpenAI API, with OpenAI planning to gradually expand access as computing power increases. OpenAI has already implemented this mode internally for incident response and plans to use speed as a new pricing lever.